utils.py

import logging
import sys
from llama_index.llms.google_genai import GoogleGenAI
from llama_index.llms.openai import OpenAI
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from sentence_transformers import CrossEncoder
from config import AVAILABLE_MODELS, DEFAULT_MODEL, GOOGLE_API_KEY
import time
from index_retriever import rerank_nodes
from utils import log_message, generate_sources_html

logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)

def log_message(message):
    logger.info(message)
    print(message, flush=True)
    sys.stdout.flush()


def get_llm_model(model_name):
    try:
        model_config = AVAILABLE_MODELS.get(model_name)
        if not model_config:
            log_message(f"Модель {model_name} не найдена, использую модель по умолчанию")
            model_config = AVAILABLE_MODELS[DEFAULT_MODEL]
        
        if not model_config.get("api_key"):
            raise Exception(f"API ключ не найден для модели {model_name}")
        
        if model_config["provider"] == "google":
            return GoogleGenAI(
                model=model_config["model_name"], 
                api_key=model_config["api_key"]
            )
        elif model_config["provider"] == "openai":
            return OpenAI(
                model=model_config["model_name"],
                api_key=model_config["api_key"]
            )
        else:
            raise Exception(f"Неподдерживаемый провайдер: {model_config['provider']}")
            
    except Exception as e:
        log_message(f"Ошибка создания модели {model_name}: {str(e)}")
        return GoogleGenAI(model="gemini-2.0-flash", api_key=GOOGLE_API_KEY)

def get_embedding_model(model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"):
    return HuggingFaceEmbedding(model_name=model_name)

def get_reranker_model(model_name='cross-encoder/ms-marco-MiniLM-L-12-v2'):
    return CrossEncoder(model_name)

def generate_sources_html(nodes, chunks_df=None):
    html = "<div style='background-color: #2d3748; color: white; padding: 20px; border-radius: 10px; max-height: 400px; overflow-y: auto;'>"
    html += "<h3 style='color: #63b3ed; margin-top: 0;'>Источники:</h3>"
    
    for i, node in enumerate(nodes):
        metadata = node.metadata if hasattr(node, 'metadata') else {}
        doc_type = metadata.get('type', 'text')
        doc_id = metadata.get('document_id', 'unknown')
        
        html += f"<div style='margin-bottom: 15px; padding: 15px; border: 1px solid #4a5568; border-radius: 8px; background-color: #1a202c;'>"
        
        if doc_type == 'text':
            html += f"<h4 style='margin: 0 0 10px 0; color: #63b3ed;'>📄 {doc_id}</h4>"
        elif doc_type == 'table':
            table_num = metadata.get('table_number', 'unknown')
            if table_num and table_num != 'unknown':
                if not table_num.startswith('№'):
                    table_num = f"№{table_num}"
                html += f"<h4 style='margin: 0 0 10px 0; color: #68d391;'>📊 Таблица {table_num} - {doc_id}</h4>"
            else:
                html += f"<h4 style='margin: 0 0 10px 0; color: #68d391;'>📊 Таблица - {doc_id}</h4>"
        elif doc_type == 'image':
            image_num = metadata.get('image_number', 'unknown')
            section = metadata.get('section', '')
            if image_num and image_num != 'unknown':
                if not str(image_num).startswith('№'):
                    image_num = f"№{image_num}"
                html += f"<h4 style='margin: 0 0 10px 0; color: #fbb6ce;'>🖼️ Изображение {image_num} - {doc_id} ({section})</h4>"
            else:
                html += f"<h4 style='margin: 0 0 10px 0; color: #fbb6ce;'>🖼️ Изображение - {doc_id} ({section})</h4>"
        
        if chunks_df is not None and 'file_link' in chunks_df.columns and doc_type == 'text':
            doc_rows = chunks_df[chunks_df['document_id'] == doc_id]
            if not doc_rows.empty:
                file_link = doc_rows.iloc[0]['file_link']
                html += f"<a href='{file_link}' target='_blank' style='color: #68d391; text-decoration: none; font-size: 14px; display: inline-block; margin-top: 10px;'>🔗 Ссылка на документ</a><br>"
        
        html += "</div>"
    
    html += "</div>"
    return html

def answer_question(question, query_engine, reranker, current_model, chunks_df=None):
    if query_engine is None:
        return "<div style='background-color: #e53e3e; color: white; padding: 20px; border-radius: 10px;'>Система не инициализирована</div>", ""
    
    try:
        log_message(f"Получен вопрос: {question}")
        log_message(f"Используется модель: {current_model}")
        start_time = time.time()
        
        log_message("Извлекаю релевантные узлы")
        retrieved_nodes = query_engine.retriever.retrieve(question)
        log_message(f"Извлечено {len(retrieved_nodes)} узлов")
        
        log_message("Применяю переранжировку")
        reranked_nodes = rerank_nodes(question, retrieved_nodes, reranker, top_k=10)
        
        log_message(f"Отправляю запрос в LLM с {len(reranked_nodes)} узлами")
        response = query_engine.query(question)
        
        end_time = time.time()
        processing_time = end_time - start_time
        
        log_message(f"Обработка завершена за {processing_time:.2f} секунд")
        
        sources_html = generate_sources_html(reranked_nodes, chunks_df)
        
        answer_with_time = f"""<div style='background-color: #2d3748; color: white; padding: 20px; border-radius: 10px; margin-bottom: 10px;'>
        <h3 style='color: #63b3ed; margin-top: 0;'>Ответ (Модель: {current_model}):</h3>
        <div style='line-height: 1.6; font-size: 16px;'>{response.response}</div>
        <div style='margin-top: 15px; padding-top: 10px; border-top: 1px solid #4a5568; font-size: 14px; color: #a0aec0;'>
        Время обработки: {processing_time:.2f} секунд
        </div>
        </div>"""
        
        return answer_with_time, sources_html
        
    except Exception as e:
        log_message(f"Ошибка обработки вопроса: {str(e)}")
        error_msg = f"<div style='background-color: #e53e3e; color: white; padding: 20px; border-radius: 10px;'>Ошибка обработки вопроса: {str(e)}</div>"
        return error_msg, ""